Information extraction method and device based on self-description network

Abstract

The invention provides an information extraction method and device based on a self-description network, and belongs to the technical field of natural languages. According to the method, based on a self-description mechanism, a unified concept set is used for representing an extraction target, association between target task data and external network data is established, and therefore the problems that annotation data is sparse and external knowledge is inaccurate are solved; generating related concepts of the extraction target by describing the generation task, taking the related concepts as descriptions of the extraction target, and constructing descriptions of the extraction type according to the descriptions of the extraction target; performing generative information extraction through an information generation task, and obtaining information such as entities and events from a text; according to the method, description generation and information generation are pre-trained by using large-scale network data, and fine tuning is performed by using a small amount of specific annotation data, so that the method can be quickly generalized to new tasks and new fields lacking annotation data.

Description

technical field [0001] The invention relates to an information extraction method and device based on a self-describing network, and belongs to the technical field of natural language processing. Background technique [0002] Information extraction aims to automatically extract structured record information from unstructured text, such record information includes but not limited to entity information, event information, etc. Taking entity information extraction as an example, given the sentence "Steve Jobs was born in San Francisco.", an information extraction system should be able to recognize that a "Steve Jobs" is a "human" entity and "San Francisco" is a "place" entity . Taking event information extraction as an example, given the sentence "Bush is due to visit the country next month.", an information extraction system should be able to identify the "visit" event, and the trigger word is "visit". Information extraction is a key task in knowledge graph construction and n...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem of sparse labeled data and mismatching external knowledge, the present invention provides an information extraction method and device based on a self-describing network, which uses a unified concept set to represent the labeled categories of various data, s

Images